Liquid subset: 500 names. Momentum/PEAD market-neutral, short-horizon reversal, and rolling-PCA residual stat-arb. Net = gross minus 10bps/side turnover cost (shifted). sqrt(365) overstates ~20% vs sqrt(252).
| Factor | Sharpe | p | maxDD | corrEqMom | corrMkt | residSharpe |
|---|---|---|---|---|---|---|
| PCAresid(K=20,hold=5,gross) | 1.23 | 0.000 | -8.9% | 0.04 | 0.02 | 0.00 |
| EqMom(long-only,net) | 0.88 | 0.000 | -55.1% | 1.00 | 0.87 | -0.00 |
| PCAresid(K=20,hold=20,gross) | 0.57 | 0.018 | -13.7% | 0.03 | 0.02 | 0.00 |
| PCAresid(K=20,hold=60,gross) | 0.36 | 0.097 | -12.5% | 0.01 | 0.02 | 0.00 |
| PCAresid(K=20,hold=60,net) | -0.01 | 0.499 | -15.5% | 0.01 | 0.01 | 0.00 |
| Rev(10d,LS,net) | -0.28 | 0.853 | -74.0% | 0.13 | 0.23 | -0.00 |
| Rev(21d,LS,net) | -0.30 | 0.864 | -76.8% | 0.01 | 0.23 | 0.00 |
| Mom(market-neutral LS,net) | -0.42 | 0.940 | -90.9% | 0.14 | -0.32 | 0.00 |
| PCAresid(K=20,hold=20,net) | -0.46 | 0.946 | -38.2% | 0.02 | 0.02 | -0.00 |
| Rev(5d,LS,net) | -0.71 | 0.995 | -92.2% | 0.18 | 0.20 | -0.00 |
| PCAresid(K=20,hold=5,net) | -2.58 | 1.000 | -89.1% | 0.03 | 0.01 | -0.00 |
| Rev(1d,LS,net) | -5.31 | 1.000 | -100.0% | 0.12 | 0.12 | -0.00 |
Reading
- Highest-Sharpe market-neutral candidates are the PCA residual stat-arb and the long-short momentum. A genuine high-Sharpe edge needs shorts (or inverse ETFs) -- the live system is long-only by design, so deployment requires a separate authorized sleeve.
- 1-day reversal is high-turnover; net Sharpe is the honest number.